2025 - Sustainable Industrial Processing Summit
SIPS2025 Volume 10. Intl. Symp on Energy, Carbon, Battery, Biochar and Agroforestry
| Editors: | F. Kongoli, S.M. Atnaw, H. Dodds, T. Turna, J. Antrekowitsch, G. Hanke, K. Aifantis, Z. Bakenov, C. Capiglia, V. Kumar, A.U.H. Qurashi, A. Tressaud, R. Yazami, M. Giorcelli |
| Publisher: | Flogen Star OUTREACH |
| Publication Year: | 2025 |
| Pages: | 316 pages |
| ISBN: | 978-1-998384-56-3 (CD) |
| ISSN: | 2291-1227 (Metals and Materials Processing in a Clean Environment Series) |
CD shopping page
MACHINE LEARNING FOR COKE QUALITY PREDICTION: ADVANCES, RESEARCH GAPS AND IMPLICATIONS FOR GHG EMISSIONS
Denilson Gomes Campos1; Paulo Assis2; Guilherme Silva3; Dimas Coura4;1FEDERAL UNIVERSITY OF OURO PRETO, Ouro Branco, Brazil; 2FEDERAL UNIVERSITY OF OURO PRETO, Ouro Preto, Brazil; 3GERDAU OURO BRANCO, Ouro Branco, Brazil; 4IFMG CAMPUS OURO, BRANCO, Conselheiro Lafaiete, Brazil;
Type of Paper: Regular
Id Paper: 353
Topic: 39
Abstract:
steel production is one of the main pillars of modern society and, although different technological routes can be used, much of the world's production depends on metallurgical coke as a strategic input for the blast furnace process. Its manufacture, however, is based on the use of coal, a non-renewable resource that is in the process of depletion. In this context, control measures and strategies for the rational use of coal are indispensable, both for the sustainability of the process and for reducing environmental impacts. The objective of this article is to conduct a bibliographic survey of publications related to machine learning techniques applied to the prediction of coke quality indices: CRI (Coke Reactivity Index), CSR (Coke Strength after Reaction), DI (Drum Index), ash, sulfur, and moisture content. The study identified a gap in research specifically focused on moisture, ash, and sulfur content indices, despite their relevance to coke quality, pointing to the need to expand studies on these parameters, which are fundamental not only to process performance but also to energy efficiency, reducing greenhouse gas (GHG) emissions, and the sustainability of steel production. In addition, this work suggests the development of coke quality prediction models based on machine learning techniques, supported by interpretability tools such as SHAP (SHapley Additive exPlanations). Complementarily, it points to the exploration of mathematical optimizations aimed at reducing the cost of the coal mix, which can be integrated with the use of genetic algorithms. These, in turn, stand out for their ability to deal with nonlinear constraints and multiple objectives, offering robust solutions for the formulation of more economical mixtures with better operational performance.